“There is huge potential for AI to change the landscape of cancer diagnosis”

What is your role at The Royal Marsden?

I am a Consultant Radiologist and I have a research appointment at The Royal Marsden as part of the AI Imaging Hub. I also work clinically in breast imaging at another trust.

I first came to The Royal Marsden as a special study module towards the end of my registrar training and was immediately impressed with the positive culture here, supporting clinicians and researchers to provide a high standard of care and enabling a wide range of research opportunities.

Can you tell me more about the Radiology and AI Research Hub?

The hub is an incredible resource, and we are so grateful for the Charity’s funding. AI research is very new, creating challenges to building expertise and understanding how best to develop and safely use this technology. The hub brings together clinicians and scientists for projects designed for patient benefit, using The Royal Marsden’s strengths in cancer imaging.  

Collaboration is the key to the hub’s success. I could not achieve anything without my colleagues, especially Des Campbell (Information Development Manager) and Donna Webb (Senior Radiographer).  

When did you first see AI being used in healthcare?

People have been trying to use AI for decades. None of the initial approaches worked very well. However, over the last 10 to 15 years, advances in computer power, algorithms, and the amount of data available have led to huge advances in AI technology, much of which has been applied to research in the healthcare space. There are still relatively few examples of such systems being used in routine care. I think this is likely to change rapidly over the next couple of years, but must be based on robust evidence for safe use. 

Can you briefly explain the AI-BRACE trial?

Artificial intelligence systems for the classification of mammography images have been developed and are beginning to be trialled and deployed for use in breast cancer screenings. Women who present with a breast cancer symptom, such as a lump, could also benefit from this technology. Mammography AI systems which have been developed in the screening setting have unknown performance in the context of symptomatic breast cancer, and may be different to screening performance - because of variations in the types and sizes of cancers presenting symptomatically, for example.

AI-BRACE will test the ability of an AI system using images and clinical data from symptomatic patients who had mammography scans over a five-year period. By using retrospective data, AI-BRACE will collate and analyse a large amount of data in a short time, helping us learn the strengths and weaknesses of the system, and how best to safely deploy this technology.

AI-BRACE will be a multi-centre study, collaborating with Imperial College Hospital NHS Trust and St George’s Hospital NHS Trust. In total around 20,000 cases will be analysed using the new AI tools. The first nine months of the project is focused on collating the images and clinical information, with testing and analysis scheduled to begin in spring 2025. We expect this to take around nine months, and then the results will be written up with the aim of publishing in summer 2026. 

What work have you been doing to prepare for the study?

One of the benefits of The Royal Marsden is that we have a superb electronic patient record (EPR). It is very accurate, complete, and also goes back an extremely long way. However, it used to be very difficult to get any information out of it at scale, because it all had to be done manually – it just was not built for analysis. 

Prior to the AI-BRACE study, Donna and I have worked closely with Des Campbell from the Performance and Innovation team, and together we have been able to curate a cohort of our moderate risk breast screening results. This is comprised of just over 5,000 patients and 26,000 attendances over a 10-year period which will form a valuable dataset for further research. Through this work we have learned techniques that we have applied to this new project and are hoping will have numerous further applications within The Royal Marsden. 

What is the potential benefit of AI-BRACE for patients?

Each year, between 3,000 and 4,000 patients attend The Royal Marsden’s symptomatic breast cancer clinic and have a mammography scan to look for cancer. If AI-BRACE and the prospective study (which hopefully will follow) are both successful, the new AI tools could potentially be rolled out for routine use at this clinic. Patients attending these clinics could therefore benefit from more accurate and efficient detection and diagnosis of their cancer. There could also be rollout of this way of working throughout the NHS, benefitting thousands more patients at other hospitals around the country every year.

What are some of the challenges when developing this project?

There are a number of challenges to be overcome, which is why the Research Hub is so important.

Testing these systems requires lots of historical information – patient records and images where we know if cancer was present or not. Although this information is recorded in patient notes, it can often only be understood by a clinical researcher. However, it is too time consuming to do this at the scale required.  In addition, some of this information (such as the exact details a patient may have presented with) comes from either unstructured fields or handwritten notes which have been scanned into the EPR. Until very recently, large-scale analysis was limited to structured fields which have been inputted electronically, meaning that it was missing a large amount of the background information from a patient’s record.

Recent advances in AI are now allowing us to develop our skills, using new tools to make more of the valuable information in our hospital records usable for testing and developing medical AI systems, while maintaining strict data security.  

What are the potential benefits of AI in the wider healthcare space?

With the research that is already taking place, as well as the rapid advancements in technology, I think there is huge potential for AI to change the landscape of cancer diagnosis. Improving patient experience is, and must always remain, the top priority when developing AI systems. Many people highlight the way in which AI can help to reduce the extreme workloads faced by healthcare staff, but this is by no means the main aim. In fact, the development of these systems and training staff to use them will be a large undertaking. Once the systems have been optimised to create the best patient experience possible, any efficiency savings will be a bonus. 

Artificial intelligence is already starting to be used for different purposes, but only in very specific and limited contexts. There is huge excitement regarding AI systems, but they need more development and testing before they can have widespread impact for patient care. 

What are your ambitions for the future of AI in healthcare?

I’d love it if, when you come to The Royal Marsden, you could benefit from the experience and evidence of the patients who have come before you, and in turn your data could be used to benefit those who come after you.